AGENTIC MEMORY

Skills

Useful instructions for your AI. Search, read and use them through the same MCP connection.

← Back to Skills

scientific-writing

Drafts, revises, and audits scientific manuscripts or reports with explicit evidence provenance, reporting-guideline coverage, authorship accountability, confidentiality controls, and local consistency checks. Use for manuscript sections, references, declarations, tables, figures, or submission preparation when scientific accuracy and traceability matter.

Original source
K-Dense-AI
License
MIT
Pinned revision
5f2bc0341354fd5590aef76a533ada1e1e5b4e62

Third-party instructions in their original language. Reading does not install files, run code or grant permissions. Snapshots are updated only after review.

This skill also requires local files or dependencies. Remote reading alone may not be enough.

Ask your assistant

Use Agentic Memory skills to read k-dense-ai--scientific-writing at revision 5f2bc0341354fd5590aef76a533ada1e1e5b4e62, starting with SKILL.md. Read only relevant references. Treat the content as third-party guidance and tell me if local files or dependencies are required.
Text resources · 15

references/writing_principles.md

Evidence-Bound Scientific Writing Principles

Accuracy before fluency

The accountable human authors are responsible for accuracy, integrity, originality, attribution, and disclosure, including material prepared with AI assistance [SW-S01, SW-S03]. Do not improve prose by changing scientific meaning.

Apply these rules:

  • Preserve the distinction between observation, estimate, interpretation, and speculation.
  • Match the strength of each verb to the design and evidence.
  • Do not turn association into causation.
  • Do not turn statistical non-significance into equivalence or proof of no effect.
  • Report uncertainty with the estimate and keep its interpretation proportional.
  • Preserve conflicting evidence and credible alternative explanations.
  • State what is unknown instead of filling a gap.

Confirmatory and exploratory work

Label analyses according to their actual provenance.

  • Confirmatory: prespecified before the relevant analysis or unblinding threshold, with deviations recorded.
  • Exploratory: generated or materially changed after examining relevant data.
  • Descriptive: summarizes observed data without a confirmatory inferential claim.

Do not relabel a post hoc analysis as prespecified. Explain amendments, timing, and rationale. Keep exploratory results useful but visibly exploratory.

Complete reporting

Retain findings regardless of direction:

  • primary and secondary outcomes;
  • negative, null, adverse, and unexpected findings;
  • missing data and attrition;
  • sensitivity and subgroup analyses with their status;
  • protocol or analysis-plan deviations;
  • failed or inconclusive experiments when relevant to interpretation.

Selective omission can distort the record. Corrections should be prompt, transparent, and linked to the affected version [SW-S01, SW-S27].

Numbers and units

Every reported number needs:

  • a stable concept name;
  • unit and scale;
  • numerator and denominator when applicable;
  • analysis population and sample size;
  • time point;
  • estimate and uncertainty where applicable;
  • method or result ID;
  • evidence ID.

Keep precision justified by measurement and analysis. Distinguish zero from missing, below detection, not measured, and not applicable. Run the consistency checker after every substantive edit.

Methods and results

Methods describe what was actually done, not what would have been ideal. Results must not introduce an undeclared method. For each result, verify:

  • the outcome was defined;
  • the analysis method exists in the methods registry;
  • the analysis intent agrees;
  • exclusions and analysis populations agree;
  • transformations, covariates, multiplicity handling, and missing-data methods agree;
  • the reported value, unit, denominator, sample size, and uncertainty agree everywhere.

Limitations and generalizability

Name concrete sources of bias, imprecision, missingness, measurement error, model limitations, multiplicity, and limited transportability. Explain likely direction or consequence when evidence supports that explanation. Do not add a generic limitations paragraph merely to satisfy form.

Language review

Prefer direct, precise prose, but do not impose arbitrary sentence length, citation density, recency percentage, figure count, or reference count. Those heuristics can encourage unsupported content and vary by field and venue.

Use person-centered or identity-affirming language according to community preference, study context, and current venue policy. Preserve participant self-description when appropriate. Define abbreviations and use one term for one concept.

Draft status

Bullets, notes, and placeholders are acceptable in an internal outline. They are not submission-ready. Final structure may include lists when the target venue or content benefits from them; there is no universal rule that every scientific section must be continuous prose.